Impact of milk preservation in the classification and prediction capabilities of a voltammetric electronic tongue

Abstract

Electronic tongues (ETs) are of great interest for the dairy industry. However, still several problems related to the fouling of the electrodes and the preservation conditions of the samples, limit their usage in milk analysis. In this work, the effects of sampling and preservation methods used to analyze raw cow milk in the performance of a portable ET based on serigraphied electrodes have been explored and optimized. Fouling caused by fat and proteins is drastically decreased if milk samples are diluted and sonicated at 20 kHz (pulses of 30 s during 5 min) previous to the analysis. Preservation by freezing at − 20ºC or by addition of azidiol have an important influence in the electrochemical responses, but freezing is a more efficient preservation method for ET analysis than azidiol. Principal Component Analysis of the signals obtained using the optimized conditions allows the discrimination of 180 milk samples with different characteristics. Using Support Vector Machine Regression (SVMR) model, the ET is able to predict the percentage of fat, protein, urea, somatic cell count (SCC) and β-Hydroxybutyric acid (BHB) with good correlation coefficients and low residual errors. The system reported can be thus used for fast in situ analysis of five parameters in a single experiment.